neighbayes.models.priors.REGibbsPriors¶
-
class neighbayes.models.priors.REGibbsPriors(beta_mu=
0.0, beta_sigma=1000000.0, rho_lower=-0.999, rho_upper=0.999, sigma2_alpha=2.0, sigma2_beta=1.0, unit_effect_sd_nu=1.0, unit_effect_sd_scale=1.0)[source]¶ Prior hyperparameters for RE panel Gibbs sampler.
Inherits
beta_mu/beta_sigma/rho_lower/rho_upperfromGibbsBasePriors. The models resolve every field to the same values their NUTS build uses, so both backends target one posterior:sigma2_alpha,sigma2_beta: σ² ~ Inv-Γ(sigma2_alpha, sigma2_beta), the models’ defaultInv-Γ(2, Var(y)).unit_effect_sd_nu,unit_effect_sd_scale: σ_α ~ half-t_ν(0, A), half-Cauchy by default (seePanelREMixinPriors), sampled through the Huang & Wand (2013) inverse-gamma mixture.
-
__init__(beta_mu=
0.0, beta_sigma=1000000.0, rho_lower=-0.999, rho_upper=0.999, sigma2_alpha=2.0, sigma2_beta=1.0, unit_effect_sd_nu=1.0, unit_effect_sd_scale=1.0)[source]¶
Methods
__init__([beta_mu, beta_sigma, rho_lower, ...])Attributes